Please mail me any criticisms, suggestions, support; offers of paper publication, lectures, and jobs, always welcome:-)


2. SUMMARY OF CRITIQUE: Outline of the critique of the PLA.
3. WORDS: Choosing suitable words to use in this unusual area. Clarifying my intended meanings.
4. POSITIVE CRITIQUE: Three kinds of putative symbolic systems, two of which - including the private one - are coherent.
5. METHODOLOGY: Criticism of Ordinary Languageism as a method for limiting language.
6. NEGATIVE CRITIQUE: Dissolving the linguistically-confused, ill-formed, questions that are usually asked, by translating them into more suitable words.
7. PASSAGES FROM THE LITERATURE: Have I missed important points? Study of some relevant texts.

Version History



A private language is self-contradictory (incoherent) because assigning private meaning to a symbol - making meaning - is impossible.


A private language needs private ostensive definition of its symbols. But this is self-contradictory, because the four criteria implicit in public ostensive definition (scene-setting, gesture, sample, projection) cannot be satisfied privately - not even if they are converted to abstracted *analogous* criteria.
In particular, the fourth criterion cannot be satisfied: Since sensations are inaccessible to others, there is no independent criterion for correctness of ascription of 'S' to a sensation; there is no method of projection, no method of checking, no normative practice, no rule. Sharability is necessary for the assignment of meaning. Without a way of checking a rule, there is no rule. There is then no distinction between correct and incorrect claims; no contentful private claims; no truth for the propositions in the 'language'.
Candlish quotes Wittgenstein (257): '"I impress [the connection] on myself" can only mean: this process brings it about that I remember the connection right in the future'. For I do not define anything, even to myself, let alone anyone else, by merely attending to something and making a mark, unless this episode has the appropriate consequences. 'Right' is not going to be the same as 'seems right to me'; but the latter, which has no evidential force, is the most I can expect to achieve in private. Before I can be said to have established a meaning for 'S', I must have set up, in advance, as part of my private system, some independent way of - some independent rule for - establishing the correctness of such later claims as "This sensation is an S". This, the fourth criterion for an ostensive definition, cannot be satisfied by a private system.
Since only the first three criteria can be satisfied, the sensation used as a sample merely *remains* the sample, available for repeated partial ostensive definition. No contentful true statement can be made, using 'S', about the sensation. For example, "That sensation is S" will always be tautological, equivalent to "That sensation is that sensation".


This Private Language Arguments start from this true premise: If a language were to be private, there would be no independent checking of putative associations between its signifier (symbol) and the signified (thing) - no first-person criterion of correctness. It proceeds to argue to the conclusion that 'private language' is inconsistent.
It purports to show the disastrous consequences of lack of checkability - to show that a private language, to be a language, must include a first-person criterion of correctness (though a public language does not need one; in it, the correctness of my propositions is regulated via the informal agreement of a network of users).
It requires, variously, linguistic essentialism combined with persistent inappropriate use of ordinary words/concepts, positivism (instrumentalism), and a verifiability criterion of meaning. Scepticism is perhaps also invoked, as Kripke and Fogelin think.
If these presuppositions are extracted, the conclusion no longer follows.

3. WORDS v1.2

1 I intentionally stated my PLA in the language of the standard literature - inappropriate everyday language. The argument is supposed to establish its conclusion conceptually. It is therefore very sensitive to meanings. Yet I made no attempt to explain my intended meaning for key words. I then used sleight of hand - persuasive definition and equivocation - to move from the vague premises, via a true result, to an exciting conclusion.

2 As the first stage of my critique, I therefore now choose meanings for some sensitive words.

3 'Languagex': This has a penumbra of meaning, partly associated with public ways of establishing meaning - by people pointing, and other people checking on correctness. Given the danger of persuasive definition, I will therefore not use it. Instead, I will use the more general concept of a 'symbolic system': A set of signifiers (symbols), signified (things, items), and associations between signifiers and signified.
'Private1': 'Not possiblyx available to others'. The sense of necessity in this description need not be further specified at this stage.
'Rulex': I will not use it. I will use 'association' and 'constant association'.
'Meaning1': The 'meaning' of a signifier is a primitive, whose meaning will be clarified, as necessary, by context. But I choose it to mean, amongst other things: 'Constant association of the signifier with a signified, or an aspect of a signified (e.g.. its redness)'. I choose not to include the requirement that there is a way of independently verifying the constancy - the correctness - of the association, when the signifier is used in propositions in later putative instances (verification by, for example, another person, or by a machine). I also choose not to include the web of associations with publicly checkable teaching-links.

4 The reader may feel very uneasy about this. She may feel that she needs to use 'languagex', 'privatex', 'rulex', and 'meaningx', with their usual meanings. In this case, she should now present her intended rough meanings for these words, for our inspection and agreement.
4.1 These words, with these meanings, are tools which are too rough for working in this delicate area. Each was devised - evolved - for an everyday use.
4.11 For example, once the usual meaning of 'language0' were presented, we might find that it included, for example, the existence of a 'rule0' for checking correctness. If so, we could deduce that a private language0 is self-contradictory - conceptually incoherent; but this would be an empty victory.
4.2 We do not accept the hegemony of ordinary language. We cannot justify insisting that unrefined ordinary words are the gold-standard for the discussion.


In a slogan: OK, a private language can't be checked. So what?

In summary: We regard the PLA as in two parts: Part 1 is correct; Part 2 is incorrect. Part 1 is the argument up to the point at which a private language is shown to be necessarily uncheckable. This is fine. But in Part 2 an exciting conclusion is deduced - a conclusion heavy with menace for traditional empiricist philosophy. This deduction is based on a tangle of confusions.

We now consider three types of symbolic system: S1, S2, and S3. We argue that type S2, of which a private language is an example, is coherent (as is type S3, our public language).

1 A symbolic system of type S1, in which there is no attempted association of signifier with signified, is contradictory, because it is not a 'system'.

2 A system of type S2, in which there is an uncheckable attempted constant association, is not contradictory.
A private system is of this type. It can operate. It does what it does - not much. This is our main result.
2.1 To presume a contradiction between 'constant association' and 'inability to check the association' is to presume a form of positivism - to presume that something that cannot be checked, is thereby non-existent. A thing whose existence has no testable consequences merely has, as Compte remarked, no place in positive, empirical, science. It is metaphysical - but this label has no automatic pejorative overtones.
2.2 We need not discuss yet whether S2 has any value. Only its coherence is in dispute.
2.21 Judgements of value are unwisely intermingled with those of coherence in some texts.
2.3 To presume that 'correct' is meaningless (nonsense) if there is no operational criterion for testing correctness is to presume the verifiability criterion of meaning.
2.4 A Robinson Crusoe system, which cannot in practice be checked, though in principle it could be (if anyone else arrives), is of this type.
2.5 A conjectural realist (constructive empiricist) accepts that this kind of system is in need of support and checking. It gets it, from the conjectured external world, and then interpersonally, later in the construction.
2.6 Symbols in S2, such as 'S', can coherently be assigned meaning(1). They can refer to mental processes; they can refer to private(1) mental events, such as particular kinds of sensing.

3 A system of type S3, in which there is checkable constant association, is not contradictory.
3.1 A public system (usually called our 'language') is of this type.
3.2 The independent checking makes S3 *more* valuable than S2. It is used for interpersonal communication, and for assisting memory.

4 None of this is surprising or significant. What is, and what is not, contradictory, is implicit in the chosen meanings of the words.

5 In system S2, how is meaning(1) made? How does the associating occur? What can an isolated person do, once he has had a sensation?
(i) He can decide that he will make the symbol for this sensation a noun, or an adjective, in a symbolic system ('Stage-setting', which is much easier in private than in public).
(ii) He can attend to the sensation (analogous to an 'ostensive gesture', and, again, much *easier* in private than in public ("Why is that man waving his arms about?") (Hacker remarks, for instance, that in the public case of 'lemony smell' there is no gesture)
(iii) and (iv): He can use his memory to decide whether a later sensation is the same as the original one (analogous to locating a 'sample' - a standard of correctness; and furthermore analogous to the 'method of projection' - laying the new entity beside the standard for comparison).
These abstracted, analogous, private versions of the criteria for public ostensive definition, are sufficient to make meaning - to achieve an attempted association of the symbol with the sensation; indeed, they are more likely to do so than analogous use of the four public criteria in S3.

6 In system S2, what can an isolated person not do - assuming that his memory is faultless?
He cannot offer an independent court of appeal to judge whether the association is being employed correctly. He has no way of checking whether he is correctly identifying a sensation as an 'S'; there is no operational distinction between him thinking that the sensation is an 'S', and the sensation actually being an 'S'.

7 Our fundamental result is this: S2 (a private language) is coherent - conceptually consistent.

8 Value: A PLA supporter could respond that the reason why he objects to S2 isn't just linguistic - that S2 isn't the kind of thing we usually call a 'language'. This, he says, is an uncharitable interpretation of his argument. He objects because a usual word such as 'language' refers to a concept of some importance - of some value. The linguistic usage has arisen because S3's cluster of features has usually been valued, and therefore named 'a language'. S2's hasn't. S2 is valueless, pointless.
8.1 This is now a judgement of value, rather than a judgement of contradictoriness. The previous claim was that a private, uncheckable, symbolic system S2 is incoherent. The new claim is that it is coherent, but valueless.
8.2 S2 is valueless, in everyday terms. But it has value to philosophers.
8.3 It has value - possibly - to an empiricist who is developing a theory concerning the structure of human knowledge. She may want to be able to refer to the private sensations. She may (arguably) want to be able to imagine an isolated Cartesian soul, presented with sensations, and conjecturing the existence of the external world, and then other minds, on the basis of these sensations.
8.31 The empiricist may find it helpful if some kind of language(x) could be devised by the isolated soul.
8.32 She does not expect that the putative associations between signifiers and signified - between 'S' and a sensation - can be checked at the initial stage of the construction.
8.33 So S2 has value to someone. That is enough.
8.4 If S2 is not of value to the empiricist either (Ayer argues that his form of empiricism relies entirely on public language) we can agree to note the coherence of S2 - and pass on to another problem.
8.5 Another philosophical value of S2 is as a rehabilitation of private meanings(1). Words, including those used in our public language, can now have three associations (related to meaning(0)):
(i) a private uncheckable reference to private sensing
(ii) a checkable public reference
(iii) public embedding in a net of teaching links.
None of these associations need be regarded as 'dominant' (to use Pears' useful word).

This completes the positive part of our critique. What is left is to unravel the confused arguments - to drain away the illusory depth - generated in our original PLA by a combination of:
(a) positivism and verificationism
(b) using inappropriate meanings extracted from everyday use, combined with:
(c) linguistic essentialism.
The first step is to reject an unjustifiable philosophical method which I call Ordinary Languageism.


1 Everyday language users, using words such as 'language', 'meaning', and 'rule', do not, in my experience, sit down over coffee to use their unexamined language to discuss the question: "Can a meaning possibly be assigned to a symbol in a private potential language?". That is not the kind of question for which everyday language has been devised. The area, if it exists, would be unusual.

2 An unusual area, with substantial issues, may exist. We may not presuppose that it does not. Philosophers may have been imagining, thinking about, and discussing, a genuine, unusual, area.
2.1 I then need to be careful to explain meanings for any everyday words which I am extrapolating into this area - and to say which other everyday words I judge not to be appropriate.
2.11 I proceed like a Physicist, who is careful to explain that 'weight' and 'amount' are going to be used with adjusted, extrapolated, meanings in his classes.
2.2 Persisting in not refining, adjusting, and extrapolating, the meanings of key words when trying to discuss the area, will lead to endless confusion and disagreement.
2.3 Denying that, for putatively unusual purposes, I can usefully adjust the meanings of words, is linguistic essentialism - the idea that each word has an essential meaning. I would then need to invent new words for all adjusted concepts. Perhaps, to reduce confusion, this should be done. Instead, I can avoid some usual words.

3 Substantial issues, in an unusual area, may not exist. They may indicate illusory depth, generated by Philosophers removing everyday words from their usual contexts, and placing them in unusual combinations; the cogs fail to mesh; language goes on holiday.
3.1 If this is so, then when I use more carefully explained words to try to express problems in the putative area, I should find that the problems dissolve; I should find that there is no unusual area - just grammatically-induced confusion.
3.2 Furthermore, if this is so, I would expect that not refining, adjusting, and extrapolating, the meanings of key words when trying to discuss the putative area, will lead to the confusion - neither locating real problems, nor dissolving them.

4 The philosopher who makes the proposal for the existence of a private language is attempting to describe ideas in an unusual area. Perhaps, (A), his proposal has only illusory depth, generated by placing usual words, with their usual meanings, in inappropriate, supposedly new, contexts - by carelessly taking usual words beyond the limits of the area they were intended for, by taking them on holiday, so that the cogs that should usually mesh to make language mean something are running free. If so, his proposal can be eliminated by careful attention to the usual uses of the words involved; examples from everyday use - presentation of facts about everyday linguistic use - will expose its nonsensical quality.
Alternatively, (B), perhaps his proposal involves many usual words, but - because of the unusual area intended - using some with adjusted meanings, and perhaps some special new words (technical terms). If so, for his critic to try to eliminate his proposal by insisting on using his words in the usual way, against his intentions, and hence triumphantly demonstrating that the philosopher's claims are inconsistent with normal usage, is unsound arguing.
4.1 I call this unsound pattern of criticism of philosophical proposals, 'Ordinary Languagism'. Its influence is pernicious.
4.2 B splits into two possibilities:
(i) His proposal is in an area which truly is beyond the limits of all language. In this case, adjusting meanings, or inventing ones, will not help him - his proposal will remain nonsense.
(ii) His proposal is beyond the limits of everyday language. It is interesting and inventive, taking human thought into unusual areas. Adjusted, or new, meanings, will be needed.
Either way, careful attention to language is essential.
4.3 I conclude that there is no method for exposing the proposal as nonsense.
4.31 The exposure of Ordinary Languagism as unsound is disappointing for a critic impatient with metaphysics - impatient to expose much historical and contemporary philosophy as nonsense he is intuitively sure it is. Any tough-minded engineer would balk at 'The Good is more identical than the Beautiful', feel that it is clearly nonsense, and feel that all we need is to find the simple method for demonstrating this. But Verificationism failed to provide an algorithm for demonstrating nonsense, and so does Ordinary Languagism.


1 "Is S2. like S3, a 'language'?" This is a merely linguistic question; it translates as: "Is S2 the kind of system which, in everyday usage, would be called a 'language'?" An appropriate, blunt, response, is: "Who cares what it would be called?". The substantial question is not "How shall we classify it?", but "What properties does it have?".
1.1 Only a person chasing the shadow of linguistic essence persists in asking: "No, but is it *really* a language?", feeling that this is an important question.

2 There is no distinction between remembering correctly the connection between 'S' and the paradigm that defined it, and merely thinking that one remembers ...if we are to talk of a 'rule' for the use of a word, then there must be an operational distinction between the correct and incorrect application of the rule.... The obeying of the rule must be a practice, exhibited in actual cases.
2.1 There is a distinction between remembering the intended association correctly, and merely thinking that one remembers. It is the distinction whose meaning we have just described in the previous sentence: in the first case we make a mistake; in the second case, we don't. This is a distinction in reality. But can I distinguish which distinct alternative is true, in practice? Can I tell if I have remembered correctly or not? I can't. That is a pity - it has already been accepted.
2.2 We are not going to talk of a 'rule', thanks very much. We are going to talk of an 'association'. The 'must' in "the obeying of a rule must be a practice, exhibited in actual cases" only gets its force from a conceptual . Our association has no independent court of appeal. A pity, but there it is.

3 If you can't describe a procedure for distinguishing two alternatives, then there aren't two alternatives. If I can't describe how I can tell if I am correctly remembering an association - a rule - rather than incorrectly, then the claim "I remembered correctly" is not potentially true or false, it is meaningless.
3.1 The first claim is positivism. The second is the verifiability criterion of meaning. Both are controversial.

4 "Rules have an essential role in language; your 'associations' are just 'rules' by another name. And rules essentially involve an operational distinction between correct and incorrect cases, exhibited in actual cases"
4.1 If we strip out the essentialism and the positivism/verificationism, this plausible sounding argument dissolves into:
(i) "Typical public symbolic systems involve putative associations where consequent correct and incorrect cases can be distinguished by public operations, exhibited in actual case." True.
(ii) "Some symbolic systems will have an independent way of distinguishing correct and incorrect cases; some won't. The latter are at a disadvantage." True.


These passages from the literature are intended:
(a) to indicate whether the statement of the PLA in these postings is typical.
(b) to indicate whether other writers are familiar with the arguments in the postings, and have familiar responses to them.

David Pears: Wittgenstein

Ludwig Wittgenstein: Philosophical Investigations

A.J.Ayer. Wittgenstein

P.M.S.Hacker: Ostensive Definition & The Private Language Argument

Stuart Candlish: The Private Language Argument


Internet References

Much of the relevant text of the Philosophical Investigations, with notes by Dr. Jones.
S.Candlish's essay on 'Private Language' is in The Stanford Encyclopedia of Philosophy

Paper References

A.J.Ayer 'Wittgenstein' 1993 Penguin (original 1985)
J.V.Canfield in R.L.Arrington & H-J. Glock ed. 'Wittgenstein's Philosophical Investigation: Text and Context' (1991) RKP
R.J.Fogelin 'Wittgenstein' (2nd Edn.) (1987) RKP
P.M.S.Hacker 'The Private Language Argument' in A Companion To Epistemology (1993) ed. J.Dancy and E.Sosa, Blackwell
D.F.Pears(1) 'Wittgenstein' (1st Edn.) 1971 Fontana/Collins
D.F.Pears(2) 'The False Prison - a Study of the Development of Wittgenstein's Philosophy' Vol.2 (1988) Clarendon Press, Oxford
L.W.Wittgenstein 'Philosophical Investigations' 1968 Edn. Blackwell (original 1953)


Version History

2nd September 1998: This essay is still in a fairly preliminary state. Obviously I haven't read all of the latest thinking - who has? Where I've used encyclopedia articles, it's because I presume that the editors reckoned that they were a definitive summary of the present state of play. I've learned a lot, in general, from the contributors to the Analytic Discussion Group on the Net, run by Rodrigo Vanegas.

Return to home page